GitHub Pauses Copilot Pro Sign-Ups and Tightens Limits Across All Individual Plans

GitHub announced on April 20 that it is pausing new sign-ups for Copilot Pro, Pro+, and Student plans. The Copilot Free tier stays open. Existing subscribers can still upgrade between paid tiers, but new accounts cannot join Pro or Pro+ until GitHub lifts the pause.

Three changes hit at once: signups frozen, usage limits tightened, and Opus models removed from the Pro tier.

Why Now

GitHub’s stated reason is straightforward: agentic workflows broke the economics of flat-rate individual plans.

When Copilot launched, the primary use case was autocomplete in an editor. A session might last an hour and use a limited amount of compute. Agentic workflows are different. A single task can run for hours, parallelize across multiple repositories, and generate far more token consumption than the original pricing assumptions built in.

From the announcement: “Agentic capabilities have expanded rapidly, agents are doing more work” and usage patterns now “regularly exceed original infrastructure assumptions, threatening service reliability for all users.”

The solution GitHub landed on was not a price increase. It was a cap and a model downgrade, at least for the lower tier.

What Changed

Usage limits: Individual plans now have two constraint types: session limits, which prevent infrastructure overload during peak demand, and weekly limits, which cap total token consumption across the billing period. Pro+ subscribers get more than five times the weekly allowance of Pro.

Starting this week, VS Code and the Copilot CLI will display warnings as users approach their thresholds, starting at 75% of the weekly limit. That context-window usage meter, which Windsurf added to Cascade earlier this month, is now coming to GitHub’s tools as well.

Model access: Opus models are gone from the Pro tier. Opus 4.7 remains available on Pro+, but Opus 4.5 and 4.6 will be discontinued from Pro+ as well. The practical effect is that Pro users are being pushed toward GPT or Sonnet-class models for agentic tasks.

One complication: Opus 4.7 carries a 7.5x premium request multiplier, meaning a single Opus session consumes tokens at 7.5x the rate of a standard request. On a plan with tighter weekly limits, that chews through an allowance fast.

Sign-up pause: Pro, Pro+, and Student plans are closed to new accounts until further notice. GitHub gave no timeline for when sign-ups would reopen.

Community Reaction

The response in GitHub’s community discussion has been pointed. The post drew 567 downvotes against 4 upvotes and 205 comments at the time of writing.

The loudest complaints center on two things: the mid-cycle model removal and annual subscribers feeling stuck.

One developer described losing a day of work: “In the middle of a very, very structured refactor, Opus 4.6 tuned into GPT…and proceeded to immediately destroy the entire day’s work.”

Others called the changes a bait-and-switch, pointing to the combination of removing features from a tier without reducing the price, then offering a short refund window as the main accommodation. Annual subscribers felt particularly exposed. The refund offer applies to monthly charges for April only, and only if users request it through GitHub Support before May 20.

What Users Can Do

GitHub’s suggestions for working within the new limits:

• Choose models with smaller request multipliers for routine tasks, saving the heavy-compute models for complex work
• Upgrade from Pro to Pro+ for the higher weekly allowance
• Use Plan mode, which drafts an action plan before executing, to reduce wasted token consumption on long tasks
• Reduce parallel workflow usage

For users who want to leave, the window is clear: request a cancellation through GitHub Support by May 20 and April charges will be waived.

GitHub’s underlying problem is one the whole industry is watching: agentic AI changes the unit economics of subscription pricing in ways that flat-rate plans can’t absorb. GitHub is not the first to adjust. It probably will not be the last.

Sources: GitHub Changelog, GitHub Blog, GitHub Community Discussion